This is the winning solution to Text Normalization Challenge - Russian Language
The project is composed of two subprojects:
- train
- translation
The first is used to train the model. Its output is a binary model file. The second is used to apply trained model to the input and do postprocessing.
Please checkout out corresponding READMEs in these projects for details on theirs execution.
These files were generated by the projects above. They can be used to jump over a few steps.
The files are in this folder on Goggle Drive
- model.t7 - trained model, you can use it to translate
- dictionary.csv - full dictionary, created according to the steps from translation/README
- restored.csv - translation output, w/o any postprocessing. You can use it to verify your training+translation procedure
- restored.pp.csv - postprocessed translation output, equal to the winning submission